噪声模糊图像的盲复原及振铃的消除
|
摘要
传统的图像复原方法往往是在假设系统的点扩散函数和噪声分布都是已知的情况下进行求解的,采用各种反卷积处理方法,如逆滤波等,对图像进行复原。然而在实际的图像处理中,许多先验知识(包括图像及成像系统的先验知识)往往并不具备,这就需要在系统点扩散函数未知的情况下,从退化图像自身求出退化信息,仅根据退化图像数据来复原真实图像,这就是盲目图像复原所要解决的问题。这是一个非常实用且富有挑战性的课题。本文的目的就是实现对未知模糊原因的模糊图像的盲复原。
图像模糊的原因主要有两个,一是点扩散函数的退化,一是噪声的引入。另外由于盲复原所得到的点扩散函数与真实点扩散函数的尺寸和数值存在偏差,所以复原的图像存在振铃效应。针对以上三个问题,本文分三个部分来解决图像盲复原问题。第一部分为噪声的滤除。为了滤除噪声和增强边缘信息,我们采用小波闽值消噪与Canny边缘检测相结合的方法。第二部分为图像的盲复原。这个部分采用迭代盲解卷积方法来进行复原。在这个过程中,首先采用零面分离法估计图像的初始值,且对原有方法的能量重新分配方案做了改进。第三部分为振铃效应的后处理。本文采用基于模糊变换理论提出的Fuzzy滤波器对振铃效应进行处理。这种方法计算量小,效果显著,是一种非常有效的解决振铃效应的方法。
图像盲复原问题本身是一个非常复杂难解的问题,在不断改进原有算法、提出新算法的同时,考虑将多种解决图像复原问题的方法结合起来将是一个新的思路。本文在盲复原前进行噪声的滤除,在盲复原后又考虑了振铃效应的处理。这样结合起来能够得到很好的复原效果,使复原质量大大提高,同时降低算法的复杂性。进一步的研究将探索如何更加有效地将各种去噪、复原以及振铃效应的后处理等方法结合起来,使复原结果更加清晰可靠。
Traditional image restoration methods are always on the assumptions that the PSF(Point Spread Function) of the system and noise distribution are known, they reconstruct images using different deconvolution approaches, for example, inverse filtering and so on. However, in real image processing, much priori knowledge (including the priori knowledge of image and the system) is not available. So it is required to determine degraded information from the degenerated image itself under the condition of unknowing the systemic PSF, and restore the real image according to the degraded image only, this constitutes the problem blind image deconvolution solving. It is a very practical and challenging subject. Our goal is to realize blind image deconvolution of blurred images whose blur information is ignorant.
There are two main reasons in image degradation, one is PSF degenerating, the other is noise. And due to the poor match between the PSF abtained by blind image deconvolution and the real PSF in size and value, there is ringing effect in deblurred image. To aim at the three problems raised above, we solve blind image deconvolution problem with three parts. Part one is denoising. To eliminate noise and enhance edge, we use the method of combining wavelet threshold denoising and canny edge detection. Part two is blind image deconvolution, the means of IBD(Iterative Blind Deconvolution)is used in this part. During this period the initial value of image is estimated by zero sheet separation approach, and the energy redistribution of original method is improved. Part three is ringing effect postprocessing. The Fuzzy filter based on the fuzzy transformation theory is employed to reduce ringing effect. Its computational complexity is low and it gains remarkable results. It is a very effective way to settle ringing effect problem.
Blind image deconvolution is a very complex and difficult topic, we should improve the existing theories constantly and propose new algorithm, in the mean while, it will be a new strategy integrating diverse technique used for image restoration. Before blind image deconvolution we make denoising, and after that we process the ringing effect. The combination can get perfect impact and increase the recovery quality remarkably, and decrease the computational complexity in the meantime. Further research to explore combining different digital image restoration technique will be a promising job.
图像模糊的原因主要有两个,一是点扩散函数的退化,一是噪声的引入。另外由于盲复原所得到的点扩散函数与真实点扩散函数的尺寸和数值存在偏差,所以复原的图像存在振铃效应。针对以上三个问题,本文分三个部分来解决图像盲复原问题。第一部分为噪声的滤除。为了滤除噪声和增强边缘信息,我们采用小波闽值消噪与Canny边缘检测相结合的方法。第二部分为图像的盲复原。这个部分采用迭代盲解卷积方法来进行复原。在这个过程中,首先采用零面分离法估计图像的初始值,且对原有方法的能量重新分配方案做了改进。第三部分为振铃效应的后处理。本文采用基于模糊变换理论提出的Fuzzy滤波器对振铃效应进行处理。这种方法计算量小,效果显著,是一种非常有效的解决振铃效应的方法。
图像盲复原问题本身是一个非常复杂难解的问题,在不断改进原有算法、提出新算法的同时,考虑将多种解决图像复原问题的方法结合起来将是一个新的思路。本文在盲复原前进行噪声的滤除,在盲复原后又考虑了振铃效应的处理。这样结合起来能够得到很好的复原效果,使复原质量大大提高,同时降低算法的复杂性。进一步的研究将探索如何更加有效地将各种去噪、复原以及振铃效应的后处理等方法结合起来,使复原结果更加清晰可靠。
Traditional image restoration methods are always on the assumptions that the PSF(Point Spread Function) of the system and noise distribution are known, they reconstruct images using different deconvolution approaches, for example, inverse filtering and so on. However, in real image processing, much priori knowledge (including the priori knowledge of image and the system) is not available. So it is required to determine degraded information from the degenerated image itself under the condition of unknowing the systemic PSF, and restore the real image according to the degraded image only, this constitutes the problem blind image deconvolution solving. It is a very practical and challenging subject. Our goal is to realize blind image deconvolution of blurred images whose blur information is ignorant.
There are two main reasons in image degradation, one is PSF degenerating, the other is noise. And due to the poor match between the PSF abtained by blind image deconvolution and the real PSF in size and value, there is ringing effect in deblurred image. To aim at the three problems raised above, we solve blind image deconvolution problem with three parts. Part one is denoising. To eliminate noise and enhance edge, we use the method of combining wavelet threshold denoising and canny edge detection. Part two is blind image deconvolution, the means of IBD(Iterative Blind Deconvolution)is used in this part. During this period the initial value of image is estimated by zero sheet separation approach, and the energy redistribution of original method is improved. Part three is ringing effect postprocessing. The Fuzzy filter based on the fuzzy transformation theory is employed to reduce ringing effect. Its computational complexity is low and it gains remarkable results. It is a very effective way to settle ringing effect problem.
Blind image deconvolution is a very complex and difficult topic, we should improve the existing theories constantly and propose new algorithm, in the mean while, it will be a new strategy integrating diverse technique used for image restoration. Before blind image deconvolution we make denoising, and after that we process the ringing effect. The combination can get perfect impact and increase the recovery quality remarkably, and decrease the computational complexity in the meantime. Further research to explore combining different digital image restoration technique will be a promising job.
引文
[1]M.R.Banham,A.K.Katasggelos.Digital image processing.IEEE signal processing magazine.1997,14(22):24-41.
[2]章毓晋.图像处理和分析(图像工程上册)[M].北京:清华大学出版社,2001.
[3]G.B.Giannakis,R.W.Heath.Blind idetification of multichannel FIR blurs and perfect image restoration.IEEE Trans.Image processing.2000,9(11):1877-1895.
[4]许峰,卢建刚,孙优贤.神经网络在图像处理中的应用.信息与控制.2003,32(4):344-351.
[5]张航,罗大庸.图像盲复原算法研究现状及展望[J].中国图象图形学报.2004,9(10):1145-1152.
[6]R.G.Lane,R.H.T.Bates.Automatic multidimensional deconvolution[J].Journal of the Optical Society.1987,4(1):180-188.
[7]FuJiwara,ZhongZhang.Defocused image restoration using translation invariant wavelet transform.SICE-ICASE Internationai joint conference,Japan,2006:3293-3297.
[8]王旭辉,郭光亚.二维匀速运动模糊图像恢复问题的研究[J].计算机应用.2000,20(10):25-28.
[9]R.L.Lagendijk,J.Biemond,D.E.Boekee.Identification and restoration of noisy blurred images using the Expectation-Maximization algorithm.IEEE Trans.Signal processing.1990,38(7):1180-1190.
[10]S.J.Reeves,R.M.Mersereau.Blur idetification by the method of generalized cross-validation.IEEE Trans.Image processing.1992,1(3):301-310.
[11]You Xu,Grey Crebbin.Image blur identification by using higher order statistic techniques.IEEE Trans.Image Processing.1996,3:77-80
[12]G.R.Ayers,J.C.Dainty.Iterative blind deconvolution method and its applications.Optics Letters.1988,13(7):547-549.
[13]B.L.K.Davey,R.G.Lane,R.H.T.Bates.Blind deconvolution of noisy complex-valued image.Optics communications.1989,69(5/6):353-356.
[14]Prashan Premaratne,Malin Premaratne.Accelerated Iterative Blind Deconvolution of Still Images image processing.IEEE TENCON 2003.Conference on Convergent Technologies for Asia-Pacific Region.2003,1:6-10.
[15]朱虹等.数字图像处理基础.北京:科学出版社,2005.
[16]飞思科技.小波分析理论与MATLAB7实现[M].北京:电子工业出版社,2005.
[17]闫传鹏.Y.Meyer型小波的构造.浙江科技学院学报.2006,18(3):177-179.
[18]Donoho D L.Denoise by softthresholding[J].IEEE Transactions on Information Theory.1995,41:613-627.
[19]赵志刚,万娇娜,管聪慧.基于小波包变换与自适应阈值的图像去噪[J].中国图象图形学报.2007,12(6):977-980.
[20]S.Grace Chang,Bin Yu,,Adaptive Wavelet Thresholding for Image Denoising and Compression[J].IEEE Trans.Image Processing.2000,9(9):1532-1546.
[21]刘守山,杨辰龙,李凌等.基于自适应小波阈值的超声信号消噪[J].浙江大学学报.2007,41(9):1557-1560.
[22]Tinku Acharya,Ajoy K.Ray著.田浩,葛秀慧,王顶等译.数字图像处理原理与应用.北京:清华大学出版社,2007.
[23]G.P.Nason,B.W.Silverman.The stationary wavelet transform and some statistical applications.Lecture Notes in Statistics,Wavelets and Statistics,Springer-Verlag,New York,1995:281-299.
[24]J.C.Pesquef,H.Krim,H.Carfantan.Time invariant orthonormal wavelet representations.IEEE Trans.On Signal Processing.1996,44(8):1964-1970.
[25]Li Xing mei,Yan Guo ping.Image denoise based on soft-threshold and edge enhancement.IEEE Digital Media and its Application in Museum&Heritages,Chongqing,2007:53-56.
[26]Canny J.A Computational Approach to Edge Detection[J].IEEE Transactions on Pattern Analysis and Machine Intelligence.1986,8(6):679-698.
[27]Kundur D,Hatzinakos D.Blind image deconvolution[J].Signal Processing Magazine.1996,13(3):43-64.
[28]胡家升.光学工程导论第二版.大连:大连理工大学出版社,2005.
[29]阮秋琦.数字图像处理学.北京:电子工业出版社,2001.
[30]J.Biemond,R.L.Lagendijk,R.M.Mersereau.Iterative methods for image deblurring.Proc.IEEE.1990,78(5):856-882
[31]A.K.Katsaggelos.Iterative image restoration algorithms.Optical Engineering.1989,28(7):735-748.
[32]Lagendijk R L,Biemond J.Iterative identification and restoration of image[M].Boston:Kluwer Academic Publishers,1991.
[33]P.Premaratne,C.C.Ko,Retrieval of symmetrical image blur using blur using zero sheets.IEEE Image and Signal Processing.2001,148(1):65-69.
[34]Bones,P.J.Parker et al.Deconvolution and phase retrieval with use of zero sheets[J].Opt.Soc.1995:1842-1857.
[35]J.Biemond,R.L.Lagendijk,R.M.Mersereau.Iterative Methods for Image Deblurring.Proceedings of the IEEE.1990,78(5):856-883.
[36]徐宗琦,高璐.一种盲复原图像振铃效应的后处理与质量评价方法.计算机应用.2007,27(4):986-988.
[37]Y.Nie,K.E.Barner.The fuzzy transformation and its applications in image processing[J].IEEE transactions on image processing.2006,15(4):910-927.
[38]K.E.Barner,Y.Nie,W.An.Fuzzy Ordering Theory and Its Use in Filter Generalization.EURASIP Journal on Applied signal Processing.2001,4:206-218
[39]A.Flaig,K.E.Barner,G.R.Arce.Fuzzy ranking:theory and applications.Signal Processing.2000,80:1017-1036.
[40] H. KONG, Y.NIE, A. YETRO et al. Coding artifacts reduction using edge map guided adaptive and fuzzy filtering[A]. IEEE ICME[C], Taipei, 2004:1135-1138.
[41] Y. Nie, K. E. Barner. Fuzzy transformation and its applications. Image Processing 2003 International Conference. 2003:1-893-896.
[2]章毓晋.图像处理和分析(图像工程上册)[M].北京:清华大学出版社,2001.
[3]G.B.Giannakis,R.W.Heath.Blind idetification of multichannel FIR blurs and perfect image restoration.IEEE Trans.Image processing.2000,9(11):1877-1895.
[4]许峰,卢建刚,孙优贤.神经网络在图像处理中的应用.信息与控制.2003,32(4):344-351.
[5]张航,罗大庸.图像盲复原算法研究现状及展望[J].中国图象图形学报.2004,9(10):1145-1152.
[6]R.G.Lane,R.H.T.Bates.Automatic multidimensional deconvolution[J].Journal of the Optical Society.1987,4(1):180-188.
[7]FuJiwara,ZhongZhang.Defocused image restoration using translation invariant wavelet transform.SICE-ICASE Internationai joint conference,Japan,2006:3293-3297.
[8]王旭辉,郭光亚.二维匀速运动模糊图像恢复问题的研究[J].计算机应用.2000,20(10):25-28.
[9]R.L.Lagendijk,J.Biemond,D.E.Boekee.Identification and restoration of noisy blurred images using the Expectation-Maximization algorithm.IEEE Trans.Signal processing.1990,38(7):1180-1190.
[10]S.J.Reeves,R.M.Mersereau.Blur idetification by the method of generalized cross-validation.IEEE Trans.Image processing.1992,1(3):301-310.
[11]You Xu,Grey Crebbin.Image blur identification by using higher order statistic techniques.IEEE Trans.Image Processing.1996,3:77-80
[12]G.R.Ayers,J.C.Dainty.Iterative blind deconvolution method and its applications.Optics Letters.1988,13(7):547-549.
[13]B.L.K.Davey,R.G.Lane,R.H.T.Bates.Blind deconvolution of noisy complex-valued image.Optics communications.1989,69(5/6):353-356.
[14]Prashan Premaratne,Malin Premaratne.Accelerated Iterative Blind Deconvolution of Still Images image processing.IEEE TENCON 2003.Conference on Convergent Technologies for Asia-Pacific Region.2003,1:6-10.
[15]朱虹等.数字图像处理基础.北京:科学出版社,2005.
[16]飞思科技.小波分析理论与MATLAB7实现[M].北京:电子工业出版社,2005.
[17]闫传鹏.Y.Meyer型小波的构造.浙江科技学院学报.2006,18(3):177-179.
[18]Donoho D L.Denoise by softthresholding[J].IEEE Transactions on Information Theory.1995,41:613-627.
[19]赵志刚,万娇娜,管聪慧.基于小波包变换与自适应阈值的图像去噪[J].中国图象图形学报.2007,12(6):977-980.
[20]S.Grace Chang,Bin Yu,,Adaptive Wavelet Thresholding for Image Denoising and Compression[J].IEEE Trans.Image Processing.2000,9(9):1532-1546.
[21]刘守山,杨辰龙,李凌等.基于自适应小波阈值的超声信号消噪[J].浙江大学学报.2007,41(9):1557-1560.
[22]Tinku Acharya,Ajoy K.Ray著.田浩,葛秀慧,王顶等译.数字图像处理原理与应用.北京:清华大学出版社,2007.
[23]G.P.Nason,B.W.Silverman.The stationary wavelet transform and some statistical applications.Lecture Notes in Statistics,Wavelets and Statistics,Springer-Verlag,New York,1995:281-299.
[24]J.C.Pesquef,H.Krim,H.Carfantan.Time invariant orthonormal wavelet representations.IEEE Trans.On Signal Processing.1996,44(8):1964-1970.
[25]Li Xing mei,Yan Guo ping.Image denoise based on soft-threshold and edge enhancement.IEEE Digital Media and its Application in Museum&Heritages,Chongqing,2007:53-56.
[26]Canny J.A Computational Approach to Edge Detection[J].IEEE Transactions on Pattern Analysis and Machine Intelligence.1986,8(6):679-698.
[27]Kundur D,Hatzinakos D.Blind image deconvolution[J].Signal Processing Magazine.1996,13(3):43-64.
[28]胡家升.光学工程导论第二版.大连:大连理工大学出版社,2005.
[29]阮秋琦.数字图像处理学.北京:电子工业出版社,2001.
[30]J.Biemond,R.L.Lagendijk,R.M.Mersereau.Iterative methods for image deblurring.Proc.IEEE.1990,78(5):856-882
[31]A.K.Katsaggelos.Iterative image restoration algorithms.Optical Engineering.1989,28(7):735-748.
[32]Lagendijk R L,Biemond J.Iterative identification and restoration of image[M].Boston:Kluwer Academic Publishers,1991.
[33]P.Premaratne,C.C.Ko,Retrieval of symmetrical image blur using blur using zero sheets.IEEE Image and Signal Processing.2001,148(1):65-69.
[34]Bones,P.J.Parker et al.Deconvolution and phase retrieval with use of zero sheets[J].Opt.Soc.1995:1842-1857.
[35]J.Biemond,R.L.Lagendijk,R.M.Mersereau.Iterative Methods for Image Deblurring.Proceedings of the IEEE.1990,78(5):856-883.
[36]徐宗琦,高璐.一种盲复原图像振铃效应的后处理与质量评价方法.计算机应用.2007,27(4):986-988.
[37]Y.Nie,K.E.Barner.The fuzzy transformation and its applications in image processing[J].IEEE transactions on image processing.2006,15(4):910-927.
[38]K.E.Barner,Y.Nie,W.An.Fuzzy Ordering Theory and Its Use in Filter Generalization.EURASIP Journal on Applied signal Processing.2001,4:206-218
[39]A.Flaig,K.E.Barner,G.R.Arce.Fuzzy ranking:theory and applications.Signal Processing.2000,80:1017-1036.
[40] H. KONG, Y.NIE, A. YETRO et al. Coding artifacts reduction using edge map guided adaptive and fuzzy filtering[A]. IEEE ICME[C], Taipei, 2004:1135-1138.
[41] Y. Nie, K. E. Barner. Fuzzy transformation and its applications. Image Processing 2003 International Conference. 2003:1-893-896.